[object Object][object Object][object Object]undefined[object Object]

• Description: The aclnnSilentCheckV2 API is added, which has different computation logic from the API. The stepRef parameter is used to determine val and compare it with the new Markov inequality threshold for verification.

• Formula:

  • If stepRef == 0:
    • If the current input [object Object] is inf/-inf/nan, or [object Object] exceeds [object Object] × [object Object], the fault is identified as an L1 fault, and an error log is printed. If the environment variable [object Object] is set to 1, [object Object] and [object Object] are updated, and then the function returns. Otherwise, inputGradRef is set to 0 and resumable training is triggered.
    • If the current input [object Object] exceeds [object Object] × [object Object], the fault is identified as an L2 fault, and a warning log is printed. [object Object] and [object Object] are updated, and then the function returns.
  • If stepRef > 0:
    • If the current input [object Object] is inf/-inf/nan, or [object Object] exceeds the Markov inequality threshold (avgRef/(1-beta1)^stepRef) × cThreshL1, the fault is identified as an L1 fault, and an error log is printed. If the environment variable [object Object] is set to 2, resumable training is triggered. If the environment variable [object Object] is set to 1, [object Object] and [object Object] are updated, and then the function returns.
    • If the current input [object Object] exceeds the Markov inequality threshold (avgRef/(1-beta1)^stepRef) × cThreshL2, the fault is identified as an L2 fault, and a warning log is printed. [object Object] and [object Object] are updated, and then the function returns.
    • If neither the L1 fault nor the L2 fault is triggered, the following occurs: If [object Object] is set to 3, the feature value of [object Object] is printed. Otherwise, [object Object] and [object Object] are updated, and then the function returns.
  • [object Object] indicates the number of detection times.

[object Object]

Each operator has calls. First, aclnnSilentCheckV2GetWorkspaceSize is called to obtain the workspace size required for computation and the executor that contains the operator computation process. Then, aclnnSilentCheckV2 is called to perform computation.

[object Object]

• Parameters:

  [object Object]

• Returns:

  [object Object]: status code. For details, see .

  The first-phase API implements input parameter verification. The following errors may be thrown:

  [object Object]

[object Object]

• Parameters:

  [object Object]

• Returns:

  [object Object]: status code. For details, see .

[object Object]

• Deterministic compute:
  • aclnnSilentCheckV2 defaults to a deterministic implementation.

[object Object]

The following example is for reference only. For details, see .